Safety of Early Rehabilitation for Acute Heart Failure Using Japanese Medical Claims

Abstract

Background

The safety of initiating early physical rehabilitation therapy and its optimal load in patients with acute decompensated heart failure (ADHF) remain unclear. This study evaluated the safe timing and load of early rehabilitation therapy for patients with ADHF.

Methods

In this cohort study using nationwide medical claims data from Japan, we included patients aged ≥18 years who were admitted for ADHF and received in‐hospital rehabilitation therapy for at least 2 days within the first 4 days of admission between 2016 and 2020. Nine hypothetical treatment regimens within the first 7 days of admission, varying in timing, duration, and load were used as exposures. The outcome was the cumulative incidence of adverse events (eg, death) within 30 days after admission. A per‐protocol analysis was conducted using a target trial emulation framework with a sequential doubly robust estimator.

Results

Altogether, 10 179 patients with ADHF were enrolled, with 752 developing adverse events (incidence 7.39%). The lowest incidence rate (4.13%) was observed in the regimen where therapy was initiated on the first day of admission with 20 minutes of therapy, continuing for up to 7 days (95% CI, 3.30–4.95). Conversely, the highest incidence rate was observed in the regimen where therapy was initiated on the third day of admission with 40 minutes of therapy, continuing for up to 7 days (8.67% [95% CI, 8.22–9.13]).

Conclusions

Very early, low‐load rehabilitation therapy for patients with ADHF may reduce the risk of adverse events. These findings suggest important implications in the implementation of early rehabilitation.

Nonstandard Abbreviations and Acronyms

ADHF

acute decompensated heart failure

DPC

Diagnosis Procedure Combination

NYHA

New York Heart Association

Clinical Perspective.

What Is New?

• This study is the first to evaluate the safety of early in‐hospital physical rehabilitation for patients with acute decompensated heart failure using a target trial emulation framework with nationwide medical claims data from Japan.

• A very early, low‐load rehabilitation regimen—initiated on the first day of admission and continued for up to 7 days—was associated with the lowest incidence of adverse events.

What Are the Clinical Implications?

• Very early, low‐intensity in‐hospital rehabilitation, initiated on the first day of admission, appeared safe even compared with naturally observed data, supporting its clinical feasibility in real‐world acute decompensated heart failure management.

The cascading health issues stemming from heart failure (HF) represent an urgent global public health concern. ¹ , ² , ³ The prevalence and incidence of HF have been rapidly increasing, particularly in countries and regions with a more pronounced aging population. ¹ Hospitalization due to acute decompensated HF (ADHF) is among the issues resulting from HF. ¹ , ² Additionally, hospitalization due to ADHF is predominantly observed in older adults ¹ , ² , ³ who often present with vulnerable conditions, including comorbidities or frailty. ⁴ , ⁵ , ⁶ , ⁷ Consequently, the safety and efficacy of in‐hospital treatment play a pivotal role in ADHF management.

A crucial element of in‐hospital care implemented in clinical practice for patients with ADHF is early physical rehabilitation therapy, which is recommended by clinical guidelines and consensus documents worldwide. ⁸ , ⁹ , ¹⁰ A recent systematic review that explored the timing of early mobilization defined early mobilization as protocol‐based physical interventions or walking within 3 days of admission. ¹¹ Nevertheless, the safety of early mobilization, that is, whether very early mobilization (initiating within 1 day of admission) is safer than moderate or late early mobilization (starting within 2–3 days after admission), remains unclear.

Safety concerns are particularly relevant in ADHF, as patients frequently present with hemodynamic instability, congestion, arrhythmias, renal impairment, or multiorgan dysfunction. ¹² These pathophysiological features may increase vulnerability to adverse events during physical activity, such as hypotension, worsening dyspnea, or falls. ¹² , ¹³ Moreover, the predominance of older and frail individuals among ADHF hospitalizations further amplifies the potential risks associated with early mobilization. ⁶ Thus, before recommending early rehabilitation as a universal strategy, its safety profile and appropriateness must be rigorously established in this acutely unstable and heterogeneous population.

When randomized clinical trials (RCTs) are unfeasible or unavailable, observational studies using routinely collected data are invaluable for evaluating the safety of management interventions. However, assessing the safety of interventions through observational studies presents several challenges. For example, older adults with ADHF receiving acute‐phase care are at a higher risk of being lost to follow‐up due to their unstable health conditions. ¹⁴ , ¹⁵ Furthermore, in clinical settings, patients receiving acute in‐hospital care are often discharged before outcome assessments can be performed. These situations are particularly problematic when evaluating the safety of interventions as such attrition can lead to an overestimation of the intervention’s safety, that is, the limitation in assuming the intention‐to‐treat principle for safety analysis. ¹⁶ , ¹⁷

To date, no study has rigorously assessed the appropriate timing and load of early physical rehabilitation therapy for patients with ADHF in terms of safety. To address this gap, we applied a causal inference‐based method and a target trial emulation framework to evaluate the safety of the timing of initiation of early in‐hospital physical rehabilitation therapy for patients with ADHF, using routinely collected data across Japan.

METHODS

Ethical approval was obtained from the Ethics Committee of the University of Occupational and Environmental Health (R4‐045). The requirement for informed consent was waived due to the anonymous nature of the data collected.

Data Availability

The data analyzed during the present study are not publicly available due to ethical reasons but are available from the corresponding author upon request.

Setting and Patients

We followed the relevant prior RCTs, review studies, and clinical guidelines on the benefits of early physical rehabilitation therapy for ADHF. ¹¹ , ¹⁸ , ¹⁹ Detailed information on the components of the target trial protocols is summarized in Table S1. We used the Japanese Diagnosis Procedure Combination (DPC) database, a case‐mix patient classification system launched in 2002 by the Ministry of Health, Labour, and Welfare of Japan. The data were collected by the DPC Study Group, a government‐funded academic organization that also gathers copies of the DPC electronic data independently of the Ministry of Health, Labour, and Welfare of Japan for research purposes. ²⁰ Detailed information about the DPC database has been previously documented. ²⁰ We analyzed the DPC data of patients meeting the hypothetical eligibility criteria between April 1, 2016 and December 31, 2020 across Japan.

Inclusion and Exclusion Criteria

The present study enrolled patients (1) aged ≥18 years who were admitted due to ADHF (International Classification of Diseases, Tenth Revision [ICD‐10] codes I50.0, I50.1, and I50.9), (2) who had not been hospitalized for ADHF within 30 days, (3) with New York Heart Association (NYHA) classification of 2 to 4, and (4) who underwent physical in‐hospital rehabilitation therapy for ≥2 days within 4 days of admission. The exclusion criteria assessed at admission were as follows: acute myocardial infarction, unstable angina pectoris, chronic obstructive pulmonary disease, anemia, active tumors, hepatic failure, chronic kidney disease (at the time of hospitalization for stage 4 or 5 and dialysis within 3 days of hospitalization), dementia, poststroke paralysis, depression, and Japan Coma Scale score of 2 to –3 digits. We further excluded patients who met the following criteria during the first 4 days after admission, to emulate the eligibility criteria (Table S1): those with cardiogenic shock; those with inotropic drug use (dopamine, dobutamine, noradrenalin, and milrinone); those using assisted circulation, such as intra‐aortic balloon pumping, percutaneous cardiopulmonary support, extracorporeal membrane oxygenation, ventilator, counter shock, extracorporeal pacemaker, amiodarone, or Shinbit use; those admitted to the intensive care unit; and those who died. These exclusion criteria were adopted to avoid violating the positivity assumption, specifically by excluding patients who were unable to start protocol‐based in‐hospital physical rehabilitation within 3 days of admission and continue it for up to 7 days.

Treatment Regimens

Generally, in‐hospital physical rehabilitation therapy programs for patients with ADHF in Japan include activities of daily living, gait, and muscle strength training. ²¹ Reimbursement for rehabilitation services >20 minutes is covered by public health insurance in Japan (1 unit). The DPC database maintains daily records of the actual cardiac rehabilitation therapy received by patients and administered by physiotherapists or occupational therapists. Considering that the median length of hospital stays for ADHF was 17 to 18 days in Japan, ²² 8.5 days in European countries, ²³ and 3 to 5 days in the United States, ²⁴ we established the following treatment regimens for up to 7 days of admission: (1) initiated on the third day of admission with 20 minutes of rehabilitation therapy (referred to as late early with low load); (2) initiated on the second day of admission with 20 minutes of rehabilitation therapy (referred to as moderate early with low load); (3) initiated on the first day of admission with 20 minutes of rehabilitation therapy (referred to as very early with low load); (4) initiated on the third day of admission with 40 minutes of rehabilitation therapy (referred to as late early with high load); (5) initiated on the second day of admission with 40 minutes of rehabilitation therapy (referred to as moderate early with high load); (6) initiated on the first day of admission with 40 minutes of rehabilitation therapy (referred to as very early and high load); (7) initiated on the third day of admission with 20 minutes of rehabilitation therapy up to the fifth day, followed by 40‐minute sessions (referred to as late early with gradually increasing load); (8) initiated on the second day of admission with 20 minutes of rehabilitation therapy up to the fourth day, followed by 40‐minute sessions (referred to as moderate early with gradually increasing load); and (9) initiated on the first day of admission with 20 minutes of rehabilitation therapy up to the third day, followed by 40‐minute sessions (referred to as very early with gradually increasing load).

Covariates

The following data were used as time‐invariant covariates: age, sex, type of HF (congestive HF, left ventricular HF, or unspecified HF), NYHA classification (2, 3, or 4), hypertension (assessed using ICD‐10 code: I10), diabetes (assessed using ICD‐10 codes: E10–E11), dyslipidemia (assessed using ICD‐10 code: E78.5), chronic kidney disease (assessed using ICD‐10 codes: N18, I12–I13), smoking history, Charlson comorbidity index (0, 1, or ≥2), admission by ambulance, Japan Coma Scale score (0 or 1), residence (own home or nursing home), home health care use before admission, body mass index (normal, underweight, overweight, or obesity), smoking status (never or former/current), Barthel Index at admission (continuous), fiscal year, and hospital volume (tertile). We also considered the following as time‐varying covariates: use of beta blocker, use of a renin–angiotensin system inhibitor, use of a mineralocorticoid receptor antagonist, use of a sodium‐glucose cotransporter 2 inhibitor, and use of intravenous nitrates. The type of HF was determined by the ICD‐10 code of the primary diagnosis with the greatest medical resource use in the DPC database.

Outcome

The outcome of this study was the occurrence of an adverse event that met the following criteria from 5 to 30 days after admission: (1) death; (2) use of inotropic drugs; (3) mechanical circulatory support, including intra‐aortic balloon pumping, percutaneous cardiopulmonary support, extracorporeal membrane oxygenation, or ventilator intubation; (4) initiation of mechanical dialysis, (5) intensive care unit readmission; or (6) readmission due to ADHF. In the present analysis, the outcome was treated as a time‐to‐event variable, and an adverse event was considered to have occurred if any of these events occurred first.

Statistical Analysis

The directed acyclic graph of this study was represented in Figure S1. The causal contrast of interest of the present study, or estimand, was the per‐protocol effect. To achieve this, the sequential doubly robust estimator was applied with a longitudinal modified treatment policy approach. This method is among the doubly robust estimations that combine both the G‐computation formula and inverse probability of the weighting model, ²⁵ , ²⁶ , ²⁷ providing unbiased estimates for counterfactual outcomes and their contrasts if either of the 2 models is consistently estimated. ²⁵ , ²⁶ , ²⁷ For more robust model specifications, we estimated both the G‐formula and inverse probability of weighting models data‐adaptively using SuperLearner, an ensemble method that employs weighted combinations of various machine learning algorithms. Generalized linear models, gradient boosting models, and neural networks were used as candidate algorithms. Using this method, the average treatment effects of each treatment regimen were calculated, along with the incidence rate, difference in the incidence rate, and relative risks. The relative risks reported in this study represent the ratios of counterfactual cumulative incidences (probabilities) of experiencing an adverse event within 30 days after admission, rather than hazard ratios (Figure S1). The reference category was set as the “late early and low load” regimen (regimen 1). In the analysis, patients discharged within 30 days of follow‐up were treated as censoring events. ²⁸ Multiple imputations were implemented to address missing values, employing the chained equation method with the random forest algorithm for imputation. ²⁹ For comparison, the incidence rate in the originally observed data was calculated, accounting for cases lost to follow‐up, and compared with the estimates for the treatment regimens (regimens 1–9). A sensitivity analysis was also performed by excluding patients admitted via ambulance, considering that nighttime admissions by ambulance could make the initiation of very early physical rehabilitation therapy on the first day of admission challenging. We accounted for clustering of patients within the same hospitals. We further calculated E‐values to assess the potential impact of unmeasured confounding, which quantify the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome to fully account for the observed associations. ³⁰ A stratified analysis by NYHA class was also performed. All analyses were conducted using R, version 4.2.2. We used the lmtp package to implement the sequential doubly robust estimator within the longitudinal modified treatment policy framework, and the SuperLearner package for data‐adaptive estimation of the G‐formula and inverse probability weighting models. Multiple imputation was performed using the mice package with the ranger package for the random forest algorithm. Candidate algorithms in the SuperLearner included the glm package for generalized linear models, the xgboost package for gradient boosting models, and the nnet package for neural networks. The detail of our analytic approach is summarized in Table S2.

RESULTS

The data of 10 179 patients (mean age was 81.2 years and 51.7% were women) who met our criteria were analyzed (Figure). Table 1 shows the patients’ baseline characteristics and health conditions. Among the patients, 46.1% were classified as having NYHA class 3. Altogether, adverse events were observed 752 cases, resulting in an incidence rate of 7.39% for time‐to‐event (Table 2). Tables S3 and S4 represented the patients’ baseline characteristics and health conditions, categorized by censoring status and the timing of initiating physical rehabilitation therapy.

Figure . Follow‐up diagram.

Table 1.

Patients’ Characteristics and Health Conditions at Baseline

Variables at admission	Number (% or mean)
No.	10 179
Age, y, mean±SD	81.2±11.4
Sex
Male	4916 (48.3%)
Female	5263 (51.7%)
Type of heart failure
Congestive heart failure	8936 (87.8%)
Left ventricular heart failure	511 (5.0%)
Unspecified heart failure	732 (7.2%)
New York Heart Association classification
2	3029 (29.8%)
3	4690 (46.1%)
4	2460 (24.2%)
Hypertension
No	4368 (42.9%)
Yes	5811 (57.1%)
Diabetes
No	7742 (76.1%)
Yes	2437 (23.9%)
Dyslipidemia
No	8721 (85.7%)
Yes	1458 (14.3%)
Chronic kidney disease
No	10 072 (98.9%)
Yes	107 (1.1%)
Charlson comorbidity index
0	4580 (45.0%)
1	3346 (32.9%)
≥2	2253 (22.1%)
Admitted by ambulance
No	7263 (71.4%)
Yes	2916 (28.6%)
Japan Coma Scale
0	9079 (89.2%)
1	1100 (10.8%)
Residence
Own home	9458 (92.9%)
Nursing home	721 (7.1%)
Home health care use before admission
No	9480 (93.1%)
Yes	664 (6.5%)
Missing	35 (0.3%)
Body mass index
Normal	5689 (55.9%)
Underweight	1385 (13.6%)
Overweight	2005 (19.7%)
Obesity	630 (6.2%)
Missing	470 (4.6%)
Smoking status
Never	6468 (63.5%)
Former/current	2819 (27.7%)
Missing	892 (8.8%)
Barthel Index at admission, mean±SD	75.5 (29.8)
Barthel Index
Dependent	3824 (37.6%)
Independent	3456 (34.0%)
Missing	2899 (28.5%)
Fiscal year
2020	605 (5.9%)
2019	2273 (22.3%)
2018	2529 (24.8%)
2017	2666 (26.2%)
2016	2106 (20.7%)
Tertile of hospital volume
Lowest	1747 (17.2%)
Middle	3368 (33.1%)
Highest	5064 (49.7%)
Beta blocker
No	7606 (74.7%)
Yes	2573 (25.3%)
Renin–angiotensin system inhibitor
No	8317 (81.7%)
Yes	1862 (18.3%)
Mineralocorticoid receptor antagonist
No	8753 (86.0%)
Yes	1426 (14.0%)
Sodium‐glucose cotransporter 2 inhibitor
No	10 093 (99.2%)
Yes	86 (0.8%)
Intravenous furosemide
No	6885 (67.6%)
Yes	3294 (32.4%)
Intravenous nitrate
No	9784 (96.1%)
Yes	395 (3.9%)

Data are expressed as the mean±SD or number (percentage). Percentages do not add up to exactly 100% because of rounding.

Table 2.

Distribution of Adverse Events in the Study Population

Type of adverse events	Number (%)
Death	199 (2.0%)
Dialysis	37 (0.4%)
Extracorporeal membrane oxygenation	0 (0.0%)
Intra‐aortic balloon pump	11 (0.1%)
Intensive care unit readmission	108 (1.1%)
Ventilator intubation	130 (1.3%)
Inotropic drug use	207 (2.0%)
Percutaneous cardiopulmonary support	0 (0.0%)
Readmission	307 (3.0%)

Table 3 shows the calculated incidence rates of adverse events at 30 days post admission for each hypothetical regimen. The incidence rate was 6.01% (95% CI, 5.56–6.47) under the late early and low load regimen, where therapy was initiated on the third day of admission with 20 minutes of therapy, continuing for up to 7 days (regimen 1). The lowest incidence rate was observed in regimen 3, where therapy was initiated on the first day of admission with 20 minutes of therapy, continuing for up to 7 days, resulting in an incidence rate of 4.13% (95% CI, 3.30–4.95). Conversely, the highest incidence rate (8.67%) was observed in regimen 4, where therapy was initiated on the third day of admission with 40 minutes of therapy, continuing for up to 7 days (95% CI, 8.22–9.13).

Table 3.

Estimated Incidence Rate of Adverse Events Among the Different Therapy Groups

Treatment regimen	Estimate	95% CIs		Difference	95% CIs		Relative risk	95% CIs		E‐value
Initiated on the third d of admission with 20 min of rehabilitation therapy, continuing for up to 7 d	6.01	5.56	6.47	Reference			Reference
Initiated on the second d of admission with 20 min of rehabilitation therapy, continuing for up to 7 d	4.50	4.15	4.84	−1.52*	−1.92*	−1.11*	0.75*	0.68*	0.83*	2.00
Initiated on the first d of admission with 20 min of rehabilitation therapy, continuing for up to 7 d	4.13	3.30	4.95	−1.88*	−2.89*	−0.88*	0.69*	0.56*	0.85*	2.26
Initiated on the third d of admission with 40 min of rehabilitation therapy, continuing for up to 7 d	8.67	8.22	9.13	2.66*	1.89*	3.44*	1.44*	1.32*	1.58*	2.24
Initiated on the second d of admission with 40 min of rehabilitation therapy, continuing for up to 7 d	5.85	5.32	6.38	−0.16	−1.00	0.68	0.97	0.86	1.10	1.21
Initiated on the first d of admission with 40 min of rehabilitation therapy, continuing for up to 7 d	5.08	3.14	7.02	−0.93	−2.94	1.07	0.85	0.57	1.25	1.63
Initiated on the third day of admission with 20 min of rehabilitation therapy up to the fifth d, followed by 40‐min sessions continuing for up to 7 d	4.59	4.45	4.74	−1.42*	−1.84*	−1.00*	0.76*	0.70*	0.83*	1.96
Initiated on the second d of admission with 20 min of rehabilitation therapy up to the fourth day, followed by 40‐min sessions continuing for up to 7 d	5.27	5.08	5.46	−0.74*	−1.19*	−0.30*	0.88*	0.81*	0.95*	1.53
Initiated on the first d of admission with 20 min of rehabilitation therapy up to the third d, followed by 40‐min sessions continuing for up to 7 d	6.01	5.31	6.70	−0.01	−1.00	0.99	1.00	0.87	1.15	1.00

All the estimates were adjusted for the following time‐invariant variables: age, sex, type of heart failure, New York Heart Association classification, hypertension, diabetes, dyslipidemia, chronic kidney disease, smoking history, Charlson comorbidity index, admission by ambulance, Japan Coma Scale, residence, home health care use before admission, body mass index, smoking status, Barthel Index at admission, fiscal year, and hospital volume. The estimates were also adjusted for the following as time‐varying variables: use of a beta blocker, use of a renin–angiotensin system inhibitor, use of mineralocorticoid receptor antagonist, use of a sodium‐glucose cotransporter 2 inhibitor, and use of intravenous nitrates. We also considered censoring weight to account for the cases lost to follow‐up.

^*Statistical significance at P<0.05.

The calculated incidence rate in the originally observed data was 6.86% (95% CI, 6.86–6.93). Compared with this rate, regimens 1–3, 5, and 7–9 were associated with a significantly lower risk of adverse events (Table 4). Additionally, regimen 4 was associated with a significantly higher risk of adverse events than those in the originally observed data (Table 4). Our sensitivity analysis, which excluded patients who were admitted by ambulance service, yielded results similar to those of the main analysis, indicating that the therapy initiated on the third day of admission was associated with a higher incidence of adverse events. Conversely, in our sensitivity analysis, very early with high‐load therapy regimens (regimens 6 and 9) showed relatively lower incidence rates of adverse events (Table S5). The results of our subgroup analysis by NYHA class are presented in Tables S6 and S7. For patients with NYHA class 2, very early regimens were generally associated with a lower risk of adverse events, and later initiation did not substantially increase the risk. In contrast, among patients with NYHA classes 3 and 4, later initiation tended to be associated with a higher risk of adverse events.

Table 4.

Comparison of the Incidence Rate of Adverse Events Between the Originally Observed Data and Hypothetical Treatment Regimens

Treatment regimen	Difference	95% CIs		Relative risk	95% CIs		E‐value
Originally observed	Reference			Reference
Initiated on the third d of admission with 20 min of rehabilitation therapy, continuing for up to 7 d	−0.85*	−1.30*	−0.40*	0.88*	0.81*	0.95*	1.53
Initiated on the second d of admission with 20 min of rehabilitation therapy, continuing for up to 7 d	−2.37*	−2.71*	−2.02*	0.66*	0.61*	0.71*	2.40
Initiated on the first d of admission with 20 min of rehabilitation therapy, continuing for up to 7 d	−2.73*	−3.56*	−1.91*	0.60*	0.49*	0.74*	2.72
Initiated on the third d of admission with 40 min of rehabilitation therapy, continuing for up to 7 d	1.81*	1.35*	2.27*	1.26*	1.20*	1.33*	1.83
Initiated on the second d of admission with 40 min of rehabilitation therapy, continuing for up to 7 d	−1.01*	−1.54*	−0.48*	0.85*	0.78*	0.93*	1.63
Initiated on the first d of admission with 40 min of rehabilitation therapy, continuing for up to 7 d	−1.78	−3.72	0.16	0.74	0.51	1.09	2.04
Initiated on the third d of admission with 20 min of rehabilitation therapy up to the fifth day, followed by 40‐min sessions continuing for up to 7 d	−2.27*	−2.42*	−2.12*	0.67*	0.65*	0.69*	2.35
Initiated on the second d of admission with 20 min of rehabilitation therapy up to the fourth day, followed by 40‐minute sessions continuing for up to 7 d	−1.59*	−1.79*	−1.40*	0.77*	0.74*	0.80*	1.92
Initiated on the first d of admission with 20 min of rehabilitation therapy up to the third day, followed by 40‐min sessions continuing for up to 7 d	−0.86*	−1.56*	−0.15*	0.88*	0.78*	0.98*	1.53

All the estimates were adjusted for the following time‐invariant variables: age, sex, type of heart failure, New York Heart Association classification, hypertension, diabetes, dyslipidemia, chronic kidney disease, smoking history, Charlson comorbidity index, admission by ambulance, Japan Coma Scale, residence, home health care use before admission, body mass index, smoking status, Barthel Index at admission, fiscal year, and hospital volume. The estimates were also adjusted for the following as time‐varying variables: use of a beta blocker, use of a renin–angiotensin system inhibitor, use of a mineralocorticoid receptor antagonist, use of a sodium‐glucose cotransporter 2 inhibitor, and use of intravenous nitrates. We also considered censoring weight to account for the cases lost to follow‐up.

^*Statistical significance at P<0.05.

DISCUSSION

The present study conducted an observational study using medical claims data from Japan to emulate an RCT and estimate the per‐protocol effect of the safety of early in‐hospital physical rehabilitation therapy for patients with ADHF. Our data showed that, among the hypothetical intervention regimens, 20 minutes of in‐hospital physical rehabilitation therapy initiated on the first day of admission and continuing for up to 7 days was associated with the lowest incidence of adverse events. Furthermore, the incidence of adverse events was lower in the regimen with only 20 minutes of therapy than in the original data.

In a previous RCT that evaluated tailored rehabilitation therapy in patients with ADHF, the patients were ambulatory before admission and were able to walk at the time of hospitalization. ¹⁸ However, in the context of ADHF, where patients are often unable to walk due to worsening hemodynamics upon admission, there are limitations in applying the findings of previous studies to clinical practice. In fact, 37% of the patients in our study were dependent on their activities of daily living at the time of admission. Moreover, a recent review has defined “early mobilization” as therapy initiated within 3 days of admission. ¹¹ However, the safety of the timing and intensity of early rehabilitation therapy remained uncertain as no RCT or observational study assessing the safety of very early rehabilitation has been conducted. ¹¹ Thus, to the best of our knowledge, the present study is the first to use a target trial emulation to evaluate safety, including patients who are often excluded from such RCTs (eg, dependent on activities of daily living before admission). ³¹ , ³² Our findings showed that very early rehabilitation, initiated on the first day of admission, with duration of 7 days and a low load of 20 minutes, was most effective in preventing adverse events within 30 days of admission. Therefore, our results suggest that low‐load rehabilitation may be beneficial during the acute phase of ADHF when implementing very early rehabilitation.

Contrarily, a regimen initiated on the third day of admission with 40 minutes of therapy, continuing for up to 7 days (regimen 4), was observed to have the highest incidence of adverse events (Table 3). A previous study has reported that introducing a high‐load rehabilitation regimen immediately after a delayed start increases the risk of adverse events, especially in vulnerable patients, which is consistent with our study findings. ³³ In fact, the late early therapy group tended to have relatively more severe health conditions (eg, NYHA classification, ambulance admission) than those in the very early and moderate early therapy groups. This indicates that our estimates for the late early and 40‐minute therapy regimens were more likely to be influenced by these severe conditions, which were adjusted for in our models (Table S4). Moreover, the results of our sensitivity analysis, excluding patients admitted by ambulance, revealed that very early rehabilitation therapy with higher‐load regimens was associated with a lower risk of adverse events (Table S8). In our sensitivity analysis, patients were more likely to have lower disease severity and higher physical functioning (Table S8). Our study data also suggest that, even if rehabilitation is delayed, a safer approach should involve gradually increasing the load from 20 to 40 minutes (Table 3). Integrating the findings of previous studies with our results highlights the importance of optimizing the intervention load for patients with ADHF according to their specific conditions and carefully considering the impact of timing and load on patient safety.

Our study findings showed that even a 20‐minute rehabilitation session, if started within 3 days of admission and continued for up to 7 days after admission, can reduce the incidence of adverse events in patients with ADHF as compared with the current real‐world rehabilitation practices (Table 4). In Japan, rehabilitation therapy for patients with ADHF generally includes activities of daily living, gait, and muscle strength training. ²¹ However, resistance training was not typically implemented in the in‐hospital setting for these patients. ²¹ Our estimations showed that very early, low‐load rehabilitation therapy led to a 0.8% to 2.7% reduction in the incidence of adverse events within 30 days of admission. Therefore, although a patient’s condition should be considered, even initiating low‐load rehabilitation therapy at a very early time point may be beneficial in actual clinical practice.

One possible mechanism underlying the effectiveness of early physical rehabilitation therapy is the gradual improvement of cardiovascular fitness and endurance. ³⁴ Regular exercise, even at low load, promotes better blood circulation and oxygen delivery to the tissues throughout the body. Early rehabilitation may also help restore the balance in the autonomic nervous system, which regulates heart rate and blood pressure, leading to improved heart rate variability and more efficient cardiovascular responses to physical exertion. ³⁵ Additionally, exercise has anti‐inflammatory effects that are beneficial to the cardiovascular system; by lowering systemic inflammation, early rehabilitation may help protect the heart and blood vessels from further damage. ³⁶ Early physical rehabilitation therapy is widely recognized to prevent disuse syndromes and complications, supporting functional maintenance. ³⁷

The population with acute HF is inherently heterogeneous, with variations in hemodynamic status, comorbidities, and functional capacity. Our additional analyses stratified by NYHA class further suggested that outcomes may differ across subgroups, highlighting the absence of a universal optimal timing or intensity of rehabilitation. Although our findings provide population‐level evidence supporting the relative safety of early rehabilitation, they should not be interpreted as a one‐size‐fits‐all recommendation. Instead, these results emphasize the importance of individualized rehabilitation strategies tailored to patient‐specific severity and risk, consistent with the established standard of individualized exercise prescription in patients with stable HF.

Our study has several limitations. First, our results may still have been influenced by unmeasured confounders (eg, the severity of systolic or diastolic HF or objectively measured functional capacity) due to the absence of such information in our data set. Moreover, although we incorporated multiple time‐varying covariates, those included in our model may not have fully captured the underlying clinical severity influencing both the initiation of rehabilitation and the risk of adverse events. However, our calculated E‐values indicated moderate robustness to unmeasured confounding (Tables 3 and 4). Furthermore, the accuracy of the data regarding the categorization of HF (ie, congestive versus left ventricular) is limited. In the DPC database, classification is determined by the ICD‐10 code of the primary diagnosis with the greatest medical resource use, which may not fully capture potential overlap between categories. Second, there may be an underestimation of the data from readmissions due to ADHF, which is one of our outcomes, because the DPC data do not include follow‐up data between hospitals. In other words, we could not obtain the data of patients who were readmitted to another hospital. Furthermore, we excluded the patients who experienced adverse events, including those receiving assisted circulation (eg, intra‐aortic balloon pumping, percutaneous cardiopulmonary support, extracorporeal membrane oxygenation), ventilator support, counter shock, extracorporeal pacemaker, amiodarone or Shinbit, those admitted to the intensive care unit, and those who died within 4 days of admission, to align with 1 of our target trial eligibility criteria, that is, “expected to be able to start protocol‐based in‐hospital physical rehabilitation within 3 days of admission, continuing up to 7 days (Table S1).” This exclusion may have led to an underestimation of the incidence of adverse events. Third, we did not consider the data of patients receiving rehabilitation therapy beyond 8 days after admission. Fourth, detailed information on rehabilitation programs was not obtained, resulting in ambiguity in the type of rehabilitation performed even when the same 20 minutes of therapy was administered. Moreover, it remains unclear whether the patients were asked to walk or perform active–passive mobilization, leaving uncertainty about the intensity of the rehabilitation and suggesting a potential consistency violation. Fifth, the proportion of missing values in the baseline Barthel Index scores (28.5%) may have introduced bias, even though we applied multiple imputation. However, the potential bias is likely to be modest, given that correlations among the individual domains were relatively high; indeed, most of the Barthel Index items included variables with R² values exceeding 0.52 (Table S9), a threshold reported to attenuate bias. ³⁸

Conclusions

In conclusion, our study findings suggested that very early and low‐load rehabilitation therapy, continuing for up to 7 days after admission, is associated with a lower risk of adverse events, including death and readmission. Our study has important implications for the implementation of early rehabilitation for patients with ADHF. However, given the heterogeneity of this population, it is essential to tailor rehabilitation strategies to patients’ specific conditions.

Sources of Funding

This work was supported by the Japan Society for the Promotion of Science [KAKENHI Grant Numbers: 22K17648 and 25K02836, Ikeda].

Disclosures

The authors declare no conflicts of interest.

Supporting information

Tables S1–S9

Figure S1

References 39,40